Researchers who wish to estimate factor demands using flexible functio
nal forms may now choose from several candidates supplied by the theor
etical literature. Unfortunately, the criteria for a priori model sele
ction are not clear. This paper adopts the use of Bayesian methods and
argues that there is in fact no need to choose; the optimal strategy
is to use a mixture of functional forms to estimate the parameters of
interest. Problems of overfitting are avoided by the imposition of the
appropriate regulatory conditions. Practical implementation is greatl
y simplified by the use of Markov chain Monte Carlo techniques. In an
example, three well-known functional forms for cost functions are appl
ied to estimate factor demand elasticities in the Canadian manufacturi
ng sector.